Spaces:

docparser

/

Text_Captcha_breaker

like 43

Runtime error

although the model is not available at the moment for public use, you can still integrate it in your application using the API endpoints

hope you consider these alternatives until the author choses to open their model for public use

Yes thank you. I am already using the api end point..

As a hobbyist would like to know how to create the model using python.. Any reference knowledge appreciated...

Thank you for your reply.

May you have a great day in Jesus Christ.

Thank you

Subin Babu

@subinbabu no need to be so formal in HF discussions, well it appears to be some sort of an OCR (optical character recognition), a pretty good one.
The only thing I can identify from the architecture is that it has a CTC head, try researching more about OCRs and how to train them.
Hopefully the author will reply to this discussion in the future providing you with more details.
Best of luck ✨✨

Sure thank you.. Will check out

You can find the open source model here. docparser/captcha

@subinbabu , please refer the above comments

@not-lain , you can use any OCR model to achieve this performance. Only the quality of the data and more augmentation are required. You can generate you own captcha images using SynthText with different configurations. Hope this helps.

@subinbabu , Please use any of these models listed for your experiments

1. PARSeq
2. ABINet
3. TRBA
4. ViTSTR-S
   5.CRNN

Thank you verymuch @docparser